Our second presentation at Isilon’s headquarters was dedicated to a closer look into Isilon’s solution and the visit of its lab.Brett Hestsel, VP of engineering gave us a more detailed presentation of Isilon’s distributed storage architecture. Brett has been with Isilon for 3 years, after working 10 years at DEC and various other high tech companies and has been in the industry for 30 years. Brett said that this is the first time in his carreer that clients come to him and tell him: ”I love your product, you made me save a lot of money!”
The real secret at Isilon is in the software, Brett explained, and the Isilon operation software is named OneFS (click here for a detailed explanation of how OneFS works). With Isilon, the whole storage is managed by software and the data is distributed across different discs within a node (aka storage server). The end result is that each node only has part of the files so that if a disc fails, all the other discs (typically 120 of them per cluster) will rebuild the data or part of the data from all the other information available from the other discs. Because there are no spare discs and all is managed via software, availability rates are much higher than more traditional storage technologies, Brett said in his presentation.
a visit to the Isilon Seattle Lab
In this lab, Isilon engineers are carrying tests on OneFS, benchmarking their solutions vs. those of competitors, testing Quality of Service (Q.O.S) and system functionality and reliability. Customers can also come to the lab and test how their applications are running in order to figure out how they run on Isilon’s storage servers.
a plug and play approach
Xavier explained that Isilon’s differentiator is that the company started from a blank sheet and didn’t have to be burdened with the existing technology and its limitations. This is has made it possible for Isilon to reinvent data storage and the way it’s managed with its distributed architecture. “Isilon applied the principles of grid computing to storage” Xavier said and this is how it works.
Each of the nodes (that is to say each rack that Xavier is pointing to in the above picture) incorporates a number of discs (typically from 12 to 36 discs) and are built in a “plug and play” fashion. This means that a rack can be added or pulled off seamlessly. “kill a node and the data will reconstruct itself across the different other nodes and disks” Xavier said. This is what makes this solution different.
a great variety of clients and industries using Isilon
The media and especially the 3D cartoon movie industry – Walt Disney namely – is relying heavily on this technology to improve performance and cost-effectiveness. Other clients include the Broad institute (life science research institute), Stratscale hosting services, Facebook , Dailymotion.com (a French contender to YouTube), Peugeot (using Isilon for crash tests), CNG (part of French nuclear research body ‘CEA’) and STMicro electronics as well as the Orange group itself (on servers operated by Orange multimedia business services, a department of Orange Business Services, on their premises)
the album of our visit to the Seattle Isilon lab